A Novel Binding Factor of 14-3-3β Functions as a Transcriptional Repressor and Promotes Anchorage-independent Growth, Tumorigenicity, and Metastasis*

The 14-3-3 proteins form a highly conserved family of dimeric proteins that interact with various signal transduction proteins and regulate cell cycle, apoptosis, stress response, and malignant transformation. We previously demonstrated that the β isoform of 14-3-3 proteins promotes tumorigenicity and angiogenesis of rat hepatoma K2 cells. In this study, to analyze the mechanism of 14-3-3β-induced malignant transformation, yeast two-hybrid screening was performed, and a novel 14-3-3β-binding factor, FBI1 (fourteen-three-three beta interactant 1), was identified. In vitro binding and co-immunoprecipitation analyses verified specific interaction of 14-3-3β with FBI1. The strong expression of FBI1 was observed in several tumor cell lines but not in non-tumor cell lines. Forced expression of antisense FBI1 in K2 cells inhibited anchorage-independent growth but had no significant effect on cell proliferation in monolayer culture. Down-regulation of FBI1 also inhibited tumorigenicity and metastasis accompanying a decrease in MMP-9 (matrix metalloproteinase-9) expression. In addition, the duration of ERK1/2 activation was curtailed in antisense FBI1-expressing K2 cells. A luciferase reporter assay revealed that the FBI1-14-3-3β complex could act as a transcriptional silencer, and MKP-1 (MAPK phosphatase-1) was one of the target genes of the FBI1-14-3-3β complex. Moreover, chromatin immunoprecipitation analysis demonstrated that FBI1 and 14-3-3β were presented on the MKP-1 promoter. These results indicate that FBI1 promotes sustained ERK1/2 activation through repression of MKP-1 transcription, resulting in promotion of tumorigenicity and metastasis.

The MAPK pathway plays an important role in diverse cellular functions, including cell proliferation, differentiation, migration, metastasis, and survival (reviewed in Refs. [12][13][14]. In response to various extracellular stimuli, extracellular signalregulated kinases (ERKs) are activated by sequential phosphorylation. Activated ERKs phosphorylate and activate downstream kinases and transcriptional regulators. Activated ERKs induce rapid transcriptional activation of immediate early genes and control the cell cycle and cell survival. Recent studies have demonstrated that differences in the duration, strength, and subcellular localization of ERK activities determined signal specificity (12,13). In addition, many studies have identified factors that may regulate the duration, strength, and subcellular localization of ERK activities. These factors contain the cell surface receptor density, the expression of scaffold proteins, and the interplay between kinases and phosphatases. Mutations in key components that result in sustained ERK activation correlate with carcinogenesis (14). For example, human tumors frequently express Ras proteins that have been activated by point mutation (15). Thus, abnormal activation of ERKs is implicated in the malignant transformation.
In this study, to further analyze the oncogenic function of 14-3-3␤, we screened K2 cell and rat brain cDNA libraries using the yeast two-hybrid system with 14-3-3␤ as a bait. One of the novel cDNA clones we isolated was designated FBI1 (fourteenthree-three beta interactant 1). We established that K2 cells down-regulated the FBI1 transcript by the introduction of an antisense FBI1 cDNA expression vector. Tumors that were formed in the flanks of nude mice by these transfectants were much smaller, and lung metastasis was robustly reduced compared with those of the parental cells. In antisense FBI1 K2 transfectants, the expression level of MMP-9 (matrix metalloproteinase-9) mRNA was reduced, whereas the expression of MKP-1 (MAPK phosphatase) was substantially up-regulated by serum stimulation. Analysis with a luciferase reporter assay using rat MKP-1 promoter and chromatin immunoprecipitation (ChIP) implied that the FBI1-14-3-3␤ complex could act as a transcriptional repressor against the MKP-1 promoter in K2 cells. Thus, these results show that the FBI1-14-3-3␤ complex plays an important role in malignant transformation through its transcriptional regulation activity.
Yeast Two-hybrid Screening-The open reading frame of rat 14-3-3␤ cDNA was inserted into pAS2-1 (Clontech) in-frame with the coding sequence of GAL4 DNA binding domain (DBD) and used for yeast two-hybrid screening as a bait vector. K2 cell and rat brain cDNA fragments were inserted into pACT2 (Clontech) in-frame with the coding sequence of the GAL4 activation domain. The FBI1 single and multiple point mutants were constructed using a QuikChange site-directed mutagenesis kit (Stratagene-Toyobo, Tokyo, Japan). Yeast twohybrid screening was performed as described previously (16).
Glutathione S-Transferase (GST) Pull-down Assay-To prepare GST-FBI1 fusion protein, an FBI1 cDNA fragment was inserted into the expression vector pGEX-2T (Amersham Biosciences) in-frame with the coding sequence of GST (pGEX-2T/FBI1). pGEX-2T/FBI1 was introduced into Escherichia coli JM109, and the fusion protein synthesized was purified on glutathione-Sepharose beads (Amersham Biosciences) as described by Kaelin et al. (17). Proteins labeled with 35 S were synthesized by the TNT coupled reticulocyte lysate system (Promega, Madison, WI). GST fusion protein (20 g) was bound to glutathione-Sepharose beads and incubated with 35 Slabeled protein (20 l) in 0.4 ml of binding buffer (phosphatebuffered saline containing 50 mM NaF, 5% glycerol, 0.5 mg/ml bovine serum albumin, and 0.3% Nonidet P-40) for 1 h at room temperature. After washing with the binding buffer, the retained proteins were subjected to 15% SDS-PAGE and visualized by autoradiography.
RNA Isolation and Northern Blot Analysis-Total RNAs were prepared from various cell lines and Fisher 344 rat tissues by the acidic guanidine thiocyanate/phenol/chloroform method. Northern blotting was performed as described previously (16).
Stable Transfections-Antisense FBI1 expression vector was constructed by the insertion of an FBI1 cDNA fragment into the pcDNA3 expression vector (pcDNA3-AS-FBI1). K2 cells were transfected with antisense FBI1 expression vector or empty vector using DOTAP transfection reagent (Roche Applied Science) according to the manufacturer's instruction. After 2 weeks of selection with 1 mg/ml G418 (Wako, Tokyo, Japan), resistant clones were expanded and analyzed for the expression level of antisense FBI1 RNA by Northern blotting.
Growth and Colonization in Vitro-Cells (4 ϫ 10 4 ) were plated in 24-well plates containing Dulbecco's modified Eagle's medium supplemented with 5% fetal calf serum (FCS) and cultivated for various times. The cell number was counted using a TATAI hematocytometer. Soft agar assay was performed as described previously (8).
Tumorigenicity and Lung Metastasis-Cells (1 ϫ 10 7 /200 l of phosphate-buffered saline/flank) were inoculated subcutaneously into 5-week-old athymic mice (BALB/c Jcl nu/nu; Clea Japan, Tokyo, Japan). After 24 days, tumor volumes were calculated as (short axis 2 ϫ long axis)/2. Thirty-one days after the injection, the lungs were removed to investigate metastatic foci. The metastatic ability was estimated by counting foci in whole mount samples. For the assay of lung metastasis, cells (1 ϫ 10 6 cells/200 l of serum-free medium) were injected into the tail vein of nude mice. After 7 weeks, metastatic foci formed in the lungs were investigated by measuring the lung weight. Mouse care and handling conformed to the National Institutes of Health guidelines for animal research. The experimental protocols were approved by the Institutional Animal Care and Use Committee.
Gelatin Zymography-When cells were grown semiconfluently, the media were changed to serum-free medium and cultivated for an additional 18 h. The resulting conditioned media (CMs) were lyophilized, resuspended in the loading buffer (10 mM Tris-HCl, pH 7.5) and electrophoresed on 10% SDS-polyacrylamide gels containing 0.1 mg/ml gelatin without denaturation. The gels were then soaked in the washing buffer (2.5% Triton X-100 in 50 mM Tris-HCl, pH 8.0) to remove SDS and incubated in the reaction buffer (50 mM Tris-HCl, pH 8.0, 10 mM CaCl 2 ) at 37°C for 18 h. The gels were subsequently stained with Coomassie Brilliant Blue and destained with 50% methanol, 10% acetic acid (v/v) to detect the gelatinase MMP-2 and MMP-9 activities.
ChIP Analysis-K2 cells were starved of serum for 12 h prior to stimulation with 10% FCS/Dulbecco's modified Eagle's medium. Various times after serum stimulation, the cells were cross-linked with 1% formaldehyde for 10 min. The nuclear fraction was isolated and sonicated to shear genomic chromatin. Chromatin extracts were precleaned with protein A-Sepharose beads at 4°C for 1 h and then incubated with 1 g of anti-FBI1, anti-14-3-3␤ (Santa Cruz Biotechnology), anti-HDAC1 (Santa Cruz Biotechnology), and anti-Sp3 (Santa Cruz Biotechnology) antibodies at 4°C overnight. Protein A-Sepharose beads were then added to the mixture for 1 h, and immunoprecipitated DNA-protein complexes were isolated from beads after several washing steps. Reversal of the cross-linking of chromatin was performed at 65°C for 6 h, and proteinase K digestion was allowed to proceed for 1 h at 55°C. DNA was extracted by the phenol/chloroform method. PCR was carried out on the purified DNA using primers corresponding to Ϫ102 to ϩ50 of the MKP-1 promoter region (5Ј-primer, 5Ј-CCG CCA TTC AAA CAA ACA AAC CG-3Ј; 3Ј-primer, 5Ј-CTG ATC CTA ATC TGG CTT CAC C-3Ј).
Statistical Analysis-All data were expressed as means Ϯ S.E. of the indicated number of experiments. Comparisons of data were carried out using analysis of variance followed by the Tukey post hoc test for multigroup comparisons or Student's t test. Differences were considered statistically significant at p Ͻ 0.05. The software package KaleidaGraph 3.6 (Synergy Software, Reading, PA) was used for statistical analysis.

RESULTS
Interaction of FBI1 with 14-3-3␤ in Vivo-In order to identify novel 14-3-3␤-interacting proteins, K2 cell cDNA and rat brain cDNA libraries were subjected to yeast two-hybrid screening using 14-3-3␤ as a bait, and three novel cDNA clones (1-1, 2-3, and 2-6) were obtained. Preliminary Northern blotting showed that the expression level of clone 2-3 mRNA in K2 cells was 24.6-fold higher than that of the normal rat liver and was the highest among these genes. Thus, in this study, we decided to further analyze the oncogenic function of clone 2-3 and designated it as FBI1. The cDNA sequence was deposited in DDBJ/ EMBI/GenBank TM under accession number AB234867. FBI1 has a sequence of 1502 nucleotides and encodes 201 amino acid residues (Fig. 1A). FBI1 protein has two putative nuclear localization signals and five putative 14-3-3 recognition sites. A homology search revealed that rat FBI1 shows 99 and 94% identities to mouse and human FBI1s at the amino acid level, respectively, indicating that FBI1 is a highly conserved protein in mammals (Fig. 1B).
To confirm the specific interaction between FBI1 and 14-3-3␤ in vitro, the GST pull-down assay was carried out. GST-fused FBI1 protein was bound to glutathione-Sepharose beads and then incubated with 35 S-labeled 14-3-3␤ synthesized by an in vitro method. As shown in Fig.  2A, the 28.2-kDa 14-3-3␤ specifically bound to GST-FBI1 fusion protein. The band detected at 26.9-kDa was perhaps due to an immature product of 14-3-3␤. To demonstrate that the interaction between these two proteins also occurs in vivo, K2 cell lysate was immunoprecipitated with anti-FBI1 antibody and subjected to Western blotting with anti-14-3-3-␤ and anti-FBI1 antibodies. As shown in Fig. 2B, endogenous 14-3-3␤ was co-immunoprecipitated with FBI1. Furthermore, FBI1 expressed in K2 cells was detected as 31.6 and 30.2 kDa bands in the relatively high molecular weight region on SDS-PAGE, probably due to its strong basicity (isoelectric point: 9.12) and the difference in phosphorylation level. FBI1 protein has five putative 14-3-3 binding sites. To verify the 14-3-3␤-binding sites in FBI1, we generated FBI1 point mutants in which Ser 31 , Thr 103 , Ser 111 , Ser 119 , or Ser 131 in the putative 14-3-3binding sites were converted to Ala and designated as S31A, T103A, S111A, S119A, and S131A, respectively. These mutants were inserted into pACT2, and the interaction with 14-3-3␤ was analyzed using the yeast two-hybrid system. All mutants did not bind to 14-3-3␤ (Fig. 2C), suggesting that multiple 14-3-3␤-binding motifs of FBI1 are required for the interaction with 14-3-3␤.
Overexpression of FBI1 mRNA in Various Tumor Cell Lines-To examine the tissue specificity of FBI1 mRNA expression, total RNAs were extracted from various adult rat tissues and analyzed by Northern blotting with the 1.3-kb fragment of pACT2/FBI1 as a probe. The FBI1 gene was markedly expressed as a 1.8-kb mRNA in the testis, cerebrum, and cerebellum, whereas expression was negligible in the liver, heart, spleen, and muscle (Fig. 3A). The reason for this difference in the expression level of FBI1 transcript among rat tissues is unclear. The FBI1 gene was robustly expressed in rat hepatoma K2 cells compared with normal rat liver. Marked expression of FBI1 mRNA was also detected in other cancer cell lines, including rat hepatomas, glioblastoma cells, pheochromocytoma cells, and embryonic carcinoma cells, compared with the respective normal tissues (Fig. 3B). On the other hand, the expression level of FBI1 in non-tumor cell lines, such as COS-7, Balb, and REF52 cells, was very low. These results suggest that overexpression of FBI1 is implicated in malignant transformation.
Specific Inhibition of Colonization in Soft Agar by Forced Expression of Antisense FBI1-It is well known that anchorageindependent growth is one of the hallmarks of malignant cancer cells. Therefore, we analyzed the effect of forced expression of antisense FBI1 on the colony forming ability of K2 cells in semisolid medium. Antisense FBI1 cDNA expression vector was introduced into K2 cells, and two stable clones, K2A1 and K2A2, were selected by Northern blotting (Fig. 3C). The expression levels of endogenous FBI1 mRNA in K2A1 and K2A2 transfectants were decreased to 64 and 65% of the parental K2 cells, respectively. The FBI1 mRNA expression level in vacant vector-introduced K2V1 cells was unaffected. Expression of the neo r gene inserted in the expression vector as a selection marker was clearly detected in all of the transfectants. Since we had previously reported that c-myc and 14-3-3␤ mRNA were over-expressed in K2 cells (8), the expression levels of these genes in the transfectants were analyzed by Northern blotting. No significant difference in the expression levels of c-myc and 14-3-3␤ mRNA was observed, suggesting that FBI1
does not affect the expression of these genes. To confirm expression level of FBI1 protein in transfectants, cell lysates were analyzed by Western blotting with anti-FBI1 antibody. The expression levels of FBI1 protein in both K2A1 and K2A2 cells were reduced to 59 and 55% compared with that of the parental K2 cells, respectively (Fig. 3D) Next, these transfectants were cultured in soft agar medium containing 10% FCS for 2 weeks, and the number of colonies was counted. The colonization abilities of K2A1 and K2A2 cells were extremely reduced to 3.5 and 11.5% compared with that of the parental K2 cells, respectively (Fig. 3, F and G). The growth rate of these antisense FBI1 transfectants in monolayer culture was similar to that of K2 cells (Fig. 3E). These results imply that FBI1 selectively promotes the anchorage-independent growth of K2 cells.
Suppression of Tumorigenicity and Lung Metastasis by Forced Expression of Antisense FBI1-The ability of anchorage-independent growth in vitro correlates with the tumorigenicity and metastatic ability of malignant cancer cells in vivo. To investigate the effects of expression of antisense FBI1 on the tumorigenicity of K2 cells, the transfectants were subcutaneously inoculated into the flank of nude mice, and the sizes of the tumors were estimated after 24 days. K2 cells and empty vectortransfected K2V1 cells formed progressively growing solid tumors in all mice (Figs. 4, A and B). Although solid tumors developed in all mice after inoculation with K2A1 and K2A2 cells, in both cases, the tumors were much smaller than those of K2 and K2V1 cells (Fig. 4, A and B). Since angiogenesis is required for vigorous tumor growth and the expansion and metastasis of solid tumors (19,20), we studied the angiogenesis of tumors derived from transfectants using anti-von Willebrand factor monoclonal antibody. The anti-von Willebrand factor-positive vessel density was not different among tumors formed by K2, K2V1, K2A1, and K2A2 cells (data not shown). K2-derived tumors in the flanks of nude mice were able to colonize the lung 1 month after subcutaneous injection. Using this assay system, we analyzed the lung metastatic potency of the antisense FBI1 transfectants. The number of lung metastatic foci developed by K2A1 and K2A2 cells was markedly diminished compared with K2 cells (Fig. 4C). Although the reduced number of foci originated from K2V1 cells was also observed, the K2V1-derived foci had a tendency to form larger foci compared with those formed by K2 cells. The reason for this is unknown. To characterize the metastatic tumors, two subcell lines, LM1 and LM2, were established from the K2A2-derived foci developed in different mice. Both cells exhibited the typical morphology of hepatocytes and were indistinguishable from the parental K2A2 cells (data not shown). Expression of the c-myc and 14-3-3␤ genes was enhanced in LM1 and LM2 cells as observed in K2A2 cells, whereas expression of these genes was negligible in the normal mouse lung (Fig. 4D). However, LM1 and LM2 cell lines acquired potent expression ability for the FBI1 gene compared with K2A2 cells, mainly as a result of the loss of antisense FBI1 RNA expression, confirmed by Northern blotting with the vector-derived BGH fragment as a probe (Fig. 4D). The expression level of FBI1 protein in LM1 and LM2 cells was higher than in K2A2 cells, coinciding with the result of Northern blotting (Fig. 4E). Furthermore, to confirm the effect of antisense FBI1 expression on metastasis, an intravenous injection assay was performed. The lungs were removed and weighed to investigate metastatic foci 7 weeks after the injection. K2 and K2V1 cells developed many metastatic foci in the lungs of nude mice, whereas no obvious nodules were observed with K2A2 transfectant. Consequently, the lung weights from the mice injected with K2A2 were robustly reduced compared with those from nude mice injected with the parental K2 cells (Fig. 4F). These results show that forced expression of antisense FBI1 inhibits the tumorigenicity and metastasis of K2 cells.
Matrix metalloproteinases (MMPs) degrade extracellular matrix, and the expression level of MMPs is correlated with the metastatic ability of cancer cells (21)(22)(23)(24). Especially, the activity of MMP-2 and -9 is often found to be elevated in tumor tissues and malignant cancer cells. Therefore, we analyzed MMP-2 and -9 activities in the CMs of K2 transfectants, LM1 and LM2, using gelatin zymography. Active MMP-9 was detected in the CMs of K2 and K2V1 cells, whereas no forms of MMP-2 were observed (Fig. 4G). In K2A1 and K2A2 CMs, active MMP-9 was slightly detected, although the pro-MMP-9 level was comparable with those of K2 and K2V1 cells. Furthermore, LM1 and LM2 cell lines acquired the strong activity of MMP-9. The activity of MMPs is controlled at both transcriptional and posttranscriptional levels (25). To elucidate the reason for the reduced level of active MMP-9 in K2A1 and K2A2 CMs, we analyzed the expression level of MMP-9 by RT-PCR. The expression level of MMP-9 mRNA in K2A1 and K2A2 cells was robustly diminished compared with levels in K2 and K2V1 cells (Fig. 4H). However, the expression levels of MMP-9 in LM1 and LM2 cell lines were equal to those in control cell lines.
Effect of Expression of Antisense FBI1 on MAPK Cascade-MAPK signaling plays a pivotal role in various cellular functions (12)(13)(14). Moreover, mutations of the MAPK pathway have been observed in many human cancers (14). Therefore, to elucidate whether overexpression of FBI1 is implicated in MAPK signal transduction, we analyzed the effect of expression of antisense FBI1 on the activation of ERK1/2 by phosphorylation. K2 cells and their antisense FBI1 transfectants were cultivated in serum-free medium for 12 h prior to stimulation with 10% serum. At various times after serum stimulation, levels of phosphorylated ERK1/2 (pERK1/2) were analyzed by Western blotting with anti-pERK1/2 antibody. In all cells tested, within 5 min, ERK1/2 were markedly phosphorylated, and the maximal levels of pERK1/2 were not significantly different among these cells. Even after 30 min, relatively high levels of pERK1/2 were sustained in K2 and K2V1 cells. In contrast, after 10 min, pERK1/2 levels were reduced immediately in K2A1 and K2A2 cells. After 30 min, pERK1/2 levels in the antisense FBI1 transfectants were reduced to their original levels. The expression levels of ERK1/2 proteins in these cells were not altered after stimulation with 10% serum (Fig. 5A).
To elucidate the reason why ERK1/2 activity was shortened in K2A1 and K2A2 cells, we analyzed the expression levels of MKP-1 by Western blotting. MKP-1 directly binds to pERK1/2 and dephosphorylated pERK1/2 (26,27). Significant induction of MKP-1 protein was observed in K2A1 and K2A2 cells after 30 -60 min of serum stimulation (Fig. 5B). In contrast, slight induction was observed in K2 and K2V1 cells. Since some MKPs are transcriptionally regulated in response to the growth factor, we analyzed the transcription levels of MKP-1 by RT-PCR. As shown in Fig. 5, C and D, the addition of serum led to induction of expression of MKP-1 mRNA shortly after activation of ERK1/2 and reached a peak level at 30 min in all cell lines tested. The relative induction levels of MKP-1 mRNA in antisense FBI1 transfectants were substantially higher than those of K2 and K2V1, although the normal expression levels of MKP-1 mRNA were different to some degree among these cells. The expression levels of MKP-2 and -3 were also analyzed by RT-PCR, and there was no significant difference among antisense FBI1 transfectants (data not shown). These results suggest that FBI1 suppresses MKP-1 gene expression. Sustained ERK activation leads to the induction of cyclin D1 and S phase entry in fibroblasts (28,29). The expression level of cyclin D1 after serum stimulation was analyzed by RT-PCR. In K2A1 and K2A2 cells, the expression of cyclin D1 was significantly reduced compared with that of K2 and K2V1 cells, suggesting that prolonged ERK activation in K2 cells may play an important role in cell cycle progression.
FBI1-14-3-3␤ Complex Has a Transcriptional Silencing Function-To examine the subcellular localization of FBI1 and 14-3-3␤ in K2 cells, immunofluorescence studies were performed. FLAG-tagged FBI1 expression vector was introduced into K2 cells, and an immunofluorescent assay was performed. FLAG-FBI1 localized in the nuclei when merged with the image of the nuclei stained with Hoechst 33258 (Fig. 6A). Moreover, enhanced green fluorescent protein-fused FBI1 expression vector was introduced into K2 cells and stained with anti-14-3-3␤ antibody. Enhanced green fluorescent protein-FBI1 was detected predominantly in the nuclei, whereas 14-3-3␤ was  Cells were cultivated in serum-free medium for 12 h and then stimulated with medium containing 10% FCS. Various times after stimulation, cell lysates and total RNAs were prepared. A, analysis of phosphorylation (left) and expression (right) levels of ERK1/2 in the transfectants by Western blotting with anti-pERK1/2 and anti-ERK1/2 antibodies, respectively. B, analysis of the serum-induced MKP-1 expression by Western blotting with anti-MKP-1 antibody (left). The expression level of actin was also analyzed as an internal control (right). C, RT-PCR analysis of the expression level of MKP-1 mRNA after serum stimulation. RT-PCR was performed using total RNAs from the serum-treated K2 transfectants. The expression level of glyceraldehyde 3Ј-phosphate dehydrogenase (GAPDH) mRNA was also analyzed as an internal control. D, the relative induction levels of MKP-1 mRNA in K2 transfectants after serum stimulation. The data obtained in C were quantitatively shown as relative expression levels. The MKP-1 mRNA level expressed in the untreated individual cells was taken as 1. E, the expression level of cyclin D1 mRNA. After 60 min of serum stimulation, total RNAs were prepared, and RT-PCR analysis was performed. The expression level of glyceraldehyde 3Ј-phosphate dehydrogenase mRNA was also analyzed as an internal control.
observed both in the cytoplasm and the nuclei (Fig. 6B). The merge of the two images showed that FBI1 and 14-3-3␤ coexisted in the nuclei. This result is supported, at least to some extent, by the accumulating evidence that 14-3-3 proteins regulate transcription in the nuclei in addition to their roles in various cellular events (30). Taking the result from immunofluorescent studies and the alteration in MMP-9 and MKP-1 mRNA expression in K2A1 and K2A2 cells as described above, we consider that FBI1 could participate in transcriptional regulation. To test this possibility, we constructed a GAL4DBDfused FBI1 expression vector, pGAL4DBD-FBI1, and the luciferase reporter plasmid pGAL4-TK-Luc containing a GAL4-binding site upstream of the TK basal promoter, and a luciferase reporter assay was performed. These plasmids were cotransfected with K2 cells in various ratios, and luciferase activity was assayed. GAL4DBD-fused FBI1 dose-dependently suppressed TK promoter activity (Fig. 7A). Since FBI1 bound with 14-3-3␤ that was overexpressed in K2 cells, we examined the effect of forced expression of antisense 14-3-3␤ on the transcriptional silencing activity of FBI1. The expression of 14-3-3␤ protein was efficiently down-regulated by the transfection with antisense 14-3-3␤, as previously reported (Fig. 7B), although some bias was observed. Coinciding with the down-regulation of 14-3-3␤, the transcriptional silencing activity of FBI1 was diminished (Fig. 7B).
Down-regulation of FBI1 expression in K2 cells altered MKP-1 mRNA expression; therefore, we hypothesized that MKP-1 is one of the targets of FBI1-14-3-3␤ complex. To analyze the effect of FBI1 and 14-3-3␤ on the MKP-1 promoter, the luciferase reporter gene plasmid Ϫ631Luc, which is driven by the region of the rat MKP-1 promoter from Ϫ631 to ϩ240, was constructed according to Ryser et al. (31) with a slight modification (Fig. 8A). The Ϫ210 to ϩ240 region as a proximal promoter contains two cAMP-responsive elements, one E box, three GC boxes, and one TATA box. K2 cells transiently transfected with Ϫ631Luc displayed substantial luciferase activities under normal conditions, and the activities were dose-dependently suppressed by the cotransfection with FBI1 expression vector (Fig. 8B). When cotransfected with 2 g of FBI1 plasmid, luciferase activity was suppressed to 46% of the control. To confirm that the silencing activity is specific for the MKP-1 promoter, luciferase assays were performed using oct3/4 and nanog promoters that contain CG boxes. FBI1 had no significant effect on these promoters (data not shown). The inhibition of Ϫ631Luc activity by 4A FBI1 mutant, which could not bind to 14-3-3␤, was hardly detected (Fig. 8C). Moreover, Ϫ631Luc activity was stimulated similarly in a dose-dependent manner by transfection with antisense 14-3-3␤ and estimated as 3.4fold of the control when cotransfected with 2 g of antisense plasmid (Fig. 8D). These results suggest that the FBI1-14-3-3␤ complex can act as a transcriptional repressor of MKP-1. To identify the region required for FBI1-14-3-3␤ repression of the MKP-1 promoter, different MKP-1 promoter constructs were introduced into K2 cells and analyzed for luciferase activities. The activity of Ϫ90Luc, which contains GC boxes and TATA box, was suppressed by FBI1 expression.
To examine whether the FBI1-14-3-3␤ complex binds to the MKP-1 promoter region in vivo, a ChIP assay was performed. In quiescent K2 cells prepared by serum starvation for 12 h, a significant amount of FBI1 and 14-3-3␤ was detected in the Ϫ102 to ϩ50 promoter region containing three GC boxes and one TATA box (Fig. 9). Interestingly, the binding level of these proteins began to decrease within 30 min after serum stimulation and reached 30% of the value for quiescent cells at 60 min. We also analyzed the binding levels of HDAC1 and Sp3, since the Sp3 transcriptional complex recruits HDAC1 and suppresses rat MKP-1 promoter activity via GC box1 (31). In a manner similar to the response profile of FBI1 and 14-3-3␤, the amount of HDAC1 and Sp3 bound to the Ϫ102 to ϩ50 promoter region in quiescent K2 cells was diminished by serum stimulation (Fig. 9).

DISCUSSION
14-3-3 proteins bind to various target proteins. Many of the target proteins (e.g. Raf-1, Bcr-Abl, TERT, and p53) act as oncogenes or tumor suppressor genes, and their functions are regulated by 14-3-3 proteins (32-35). 14-3-3 is well known as a tumor suppressor gene; in many kinds of cancer tissues, 14-3-3 is down-regulated by promoter methylation or degradation (36,37). On the other hand, several studies have reported that some isoforms of 14-3-3 are overexpressed in specific cancer cells and have provided some correlative information between 14-3-3 proteins and oncogenesis (38 -40). However, there are few reports that demonstrate the molecular mechanisms of 14-3-3 in oncogenesis. We previously reported that 14-3-3␤ is overexpressed in various cancer cells and that down-regulation of 14-3-3␤ proteins results in suppression of anchorage-independent growth, tumorigenicity, and angiogenesis of rat hepatoma K2 cells (8). Thus, 14-3-3␤ may act as an oncogene. In this study, we attempted to demonstrate the mechanisms of 14-3-3␤induced cell growth and tumorigenesis by screening binding partners. We identified FBI1 as a novel binding partner of 14-3-3␤ and FBI1 promoted anchorage-independent growth, tumorigenicity, and metastasis. In addition, luciferase reporter and ChIP assays revealed that 14-3-3␤ and FBI1 formed a repressor complex on the MKP-1 promoter.
The immunoprecipitation and pull-down analyses demonstrated that FBI1 binds to 14-3-3␤ directly in vitro and in vivo. Luciferase assay showed that multiple mutations of binding sites were required to completely abrogate the silencing activity of FBI1. This result suggests FBI1 binds to 14-3-3␤ through multiple sites. 14-3-3 proteins are dimeric proteins; thus, they can interact with a single FBI1 protein as well as Raf-1, Bad, and Cdc25B, which also contain some 14-3-3 binding sites (41)(42)(43). Cdc25B contains three 14-3-3 binding sites. Although single mutations of these binding sites reduce the interaction with FIGURE 8. The FBI1-14-3-3␤ complex suppresses MKP-1 promoter activity. A, schematic illustration of the rat MKP-1 promoter-linked luciferase (Luc) reporter plasmids Ϫ631Luc, Ϫ210Luc, Ϫ114Luc, and Ϫ90Luc. B, effects of FBI1 expression on the MKP-1 promoter activity. The reporter plasmid (0.25 g) and the ␤-galactosidase expression plasmid (0.25 g) were introduced into K2 cells together with total 2.0 g of pcDNA3 and pcDNA3-FBI1 in various combinations. Luciferase activities were analyzed as described above. C, FBI1-14-3-3␤ complex formation is required for their transcriptional silencing activity. The 4A multiple point mutant of pFLAG-CMV2-FBI1 expression vector was introduced into K2 cells in the same conditions described above and assayed for luciferase activity. D, effects of the down-regulation of 14-3-3␤ on the transcriptional silencing activity. pcDNA3-AS-14-3-3␤ was introduced into K2 cells as described above and assayed for luciferase activity. E, identification of the FBI1 response element of the MKP-1 promoter. Different MKP-1 promoter constructs were introduced into K2 cells together with total 2.0 g of pcDNA3 and pcDNA3-FBI1 in various combinations. Luciferase activities were analyzed as described above. Each value indicates the mean Ϯ S.E. of triplicate assays. *, p Ͻ 0.01 compared with the K2 cells transfected with only Ϫ630Luc (B-E).
14-3-3, these mutations are not sufficient to inhibit the interaction and the functional effect of 14-3-3 on Cdc25B activity (44). Likewise, FBI1 single mutants can reduce but not abolish the binding to 14-3-3␤. Our yeast two-hybrid system probably could not detect the reduced binding. Therefore, all yeast transformed with single mutations of FBI1 did not form colonies. Cdc25B is phosphorylated at serine residues that are adjacent to 14-3-3 binding sites of the N terminus. These phosphorylations regulate interaction between Cdc25B and 14-3-3 (45). Bad also has additional phosphorylation sites that regulate interaction with 14-3-3 (43). 14-3-3 binding sites of FBI1, Thr 103 , Ser 111 , and Ser 119 , are located close to each other. In addition, Western blot analysis revealed that FBI1 protein was detected as some bands, suggesting the possibility of phosphorylation. Therefore, it is likely that multiple phosphorylations of FBI1 are required for stable and functional complex formation with 14-3-3␤.
Our studies showed that high expression of FBI1 in K2 cells promotes anchorage-independent growth, tumorigenicity, and metastasis. The strong induction of MKP-1 mRNA and protein was observed in K2 cells expressing antisense FBI1. In addition, the duration of ERK1/2 activation was shortened, indicating that up-regulated MKP-1 probably inactivates ERK1/2 in antisense FBI1 transfectants. Recent studies showed that differences in the duration of ERK1/2 activity determine signal specificity (12,13). In fibroblast cells, sustained activation of ERK1/2 is required for the stabilization of c-Fos proteins and the expression of cyclin D1 that promotes G 1 /S transition (28,29). In antisense transfectants, the decrease in cyclin D1 expression was also observed. These results suggest that FBI1 can promote cell growth by extending the duration of ERK1/2 activation. During metastasis, cancer cells must cross several extracellular matrix barriers. MMPs are activated in several stages of metastasis and degrade extracellular matrix components (21,25). MMP-9 degrades type IV collagen, which constitutes the major component of the basement membrane and promotes metastasis (22,23), indicating that the activity of MMP-9 plays an important role in metastasis. The expression of MMP-9 was reduced in antisense FBI1 transfectants. However, in metastatic cell lines, which acquired high expression of FBI1 protein, the expression of MMP-9 was comparable with that in K2 cells. It is possible that FBI1 may regulate the expression level of MMP-9 and promote metastasis.
Luciferase reporter and ChIP analyses revealed that the FBI1-14-3-3␤ complex could act as a transcriptional silencer and that MKP-1 is one of the target genes. Several reports have demonstrated that 14-3-3 proteins regulate transcription. In most cases, the interaction between 14-3-3 proteins and transcriptional factors leads to sequestration in the cytoplasm, away from their transcriptional targets (30,46). However, our studies revealed that FBI1 formed a complex with 14-3-3␤ on the MKP-1 promoter and that the complex suppressed MKP-1 transcription in K2 cells. Thus, these results propose a novel function of 14-3-3 in transcriptional regulation. Luciferase reporter analysis using different MKP-1 promoter constructs showed that the region from Ϫ90 to ϩ240 containing GC boxes and a TATA box is important for suppression of the FBI1-14-3-3␤ complex. In addition, Sp3 and HDAC1 bound to the MKP-1 promoter in the absence of serum stimulation. Sp3, which can recruit HDAC1/2, suppresses MKP-1 promoter activity via the GC box1 (GGACCGCCCC) in the Ϫ203 to ϩ19 proximal promoter region (31). It is possible that the FBI1-14-3-3␤ complex suppresses MKP-1 promoter activity via the GC box 1 in cooperation with the Sp3 transcriptional repressor complex containing HDAC1.
There are at least seven isoforms of 14-3-3 in mammals. 14-3-3 is also highly expressed in some cancer tissues (38). A recent report has demonstrated that knockdown of 14-3-3 in a lung cancer cell line induces anoikis (47). The anoikis induction is mediated by reduction of Akt activity and induction of BH3only proapoptotic proteins. On the other hand, in our studies, the activity of Akt was not detected in K2 cells, and there was no significant difference in apoptosis induction by some apoptotic inducers (data not shown). Therefore, 14-3-3␤ induces cell growth, tumorigenicity, and metastasis by different molecular mechanisms from that of 14-3-3, suggesting that 14-3-3␤ is a unique isoform among 14-3-3 family proteins.
We demonstrated that FBI1 plays a pivotal role in metastasis of K2 cells as well as tumor growth. Metastasis is the cause of 90% of human cancer deaths. However, little is known about its genetic and biochemical determinants. Since FBI1 was overexpressed in several cancer cell lines, identification of FBI1 target genes and investigation of the molecular mechanism of FBI1 transcriptional regulation may help in further understanding the mechanism of metastasis and in the search for new therapeutic targets. FIGURE 9. ChIP analysis of MKP-1 promoter. K2 cells were cultivated in serum-free medium for 12 h and then stimulated by 10% FCS. After 30 and 60 min, the cells were cross-linked with 1% formaldehyde, and the soluble chromatin fraction was prepared by sonication and immunoprecipitated with anti-FBI1, anti-14-3-3␤, anti-HDAC1, and anti-Sp3 antibodies. DNA was extracted from the immunoprecipitates and analyzed by PCR using primer pairs covering the Ϫ102 to ϩ50 MKP-1 promoter region containing three GC boxes and one TATA box. The forward and reverse primer positions are shown schematically by arrows in the top panel.